| Blood cell signal (BCS) is a kind of signals which has various shapes of pulseand time-frequency features. It is difficult to embody the physiological andpathological information of BCS accurately by using cell signal pulse-countingmethod in clinics since the BCS is of nonlinear, non-stationary and non-gaussianfeatures. In addition, it is hard to keep the good performance of cell counting becauseof the influence of some factors，like the M signal. The cell signal pulse-countingmethod may be not efficient and reliable.In order to improve the situation, the adaptive HHT analysis method with ahigh frequency concentration is proposed to process the BCS in this dissertation. Thefeature vectors are extracted for BCS classification which use the support vectormachine (SVM). Firstly, the BCS is decomposed into several intrinsic mode function(IMF) components by empirical mode decomposition (EMD) method adaptively. Andthe Hilbert spectrum, Hilbert marginal spectrum and Fourier spectrum are obtained.Secondly, the time-frequency feature and nonlinear dynamics feature of BCS isanalyzed. The average intensity, spectral centric, the energy contribution rate andmarginal spectrum entropy of the healthy people and patients’ BCS are extractedrespectively. The time-frequency feature vectors and nonlinear dynamics featurevectors are constructed for classification. Finally, the experiments are carried out totest the effect of HHT method in processing BCS.The results show that the features extracted by HHT method are discriminativein the classification of BCS of healthy people and patients. The highest classificationrate (CR) of time-frequency feature vectors is94.12%. And the highest CR ofnonlinear dynamics feature vectors is92.16%. HHT method is significative to assistBCS clinical analysis and processing. It can provide a new way for the application ofBCS analysis. |
